NezhaGateNezhaGate
AnnouncementsAll updates →
NoticeWan 3.0: Standard 480P / 720P / 1080P available; Prime paused NewKling 4.0 is live: reference images, videos and audio, 3-30 seconds, billed per second at 720P / 1080P NoticeFrom October 29: top-up bonuses by payment method and a $1.40 refund fee
简体中文English ✓Español한국어日本語Русский
Log in Get API Key
Status

Service status

Live status of every model, a daily record for the last 90 days, and the incident history. Every figure comes from real customer requests; internal tests are excluded.

5 models affected

The other 36 models are operating normally.


Refreshes every minute

All services

Customer request success rate
93.6%Last 24 hours
95.5%Last 90 days
90 days ago30 days agoToday
OperationalDegradedOutageMaintenanceNo data

Trends

Daily success rate

Per customer request · UTC days

Latency

Streaming requests, from sending to the first token (thinking included)

Medianp90

Models 90-day success rate

Chat models

DeepSeek V4.1 Flash
Degradedelevated error rate
97.3%
Gemini 3.1 Pro
MaintenanceTemporarily suspended. The model listing and documentation are retained; reopening plans will be announced separately
99.7%
GLM-5.3
99.7%
GPT-5.6 Luna
97.7%
GPT-5.6 Sol
97.7%
GPT-6 Astra
94.4%
GPT-6 Luna
99.0%
GPT-6 Sol
99.5%
GPT-6.1 Sol
95.0%
Grok 4.7
91.6%
Qwen3.8 Flash
Degradedanswered mostly by another model
99.1%

Image models

GPT Image 2
98.7%

Video models

MiniMax H3
90.0%
Seedance 2.0
62.8%
Seedance 2.0 · Clip
Maintenanceline maintenance; we will announce when it is back
Seedance 2.5
Outagemost requests failing
78.2%
Wan 3.0
60.7%

Incident history

Outage

Seedance 2.5: most requests failing

Ongoing
– Ongoing · ongoing for 1 h · Detected automatically

Failed requests rose noticeably; failed requests are not billed.

Degraded

DeepSeek V4.1 Flash: elevated error rate

Ongoing
– Ongoing · ongoing for 3 h 56 min · Detected automatically

Failed requests rose noticeably; failed requests are not billed. Success rate during the incident: 75.8%.

Maintenance

Gemini 3.1 Pro: Maintenance

Ongoing
– Ongoing · ongoing for 4 d 16 h

Temporarily suspended. The model listing and documentation are retained; reopening plans will be announced separately During maintenance the API answers “temporarily unavailable” right away; nothing is billed and no other model answers in its place.

Degraded

Qwen3.8 Flash: answered mostly by another model

Ongoing
– Ongoing · ongoing for 9 d 5 h · Detected automatically

Every request was answered, but 100% of them by another model.

Maintenance

Seedance 2.0 · Clip: Maintenance

Ongoing
– Ongoing · ongoing for 16 d 13 h

line maintenance; we will announce when it is back During maintenance the API answers “temporarily unavailable” right away; nothing is billed and no other model answers in its place.

Degraded

GPT Image 2.5 Sunburst: generation noticeably slower

– · lasted 50 min · Detected automatically

Response times ran well above this model's previous 7 days. Success rate during the incident: 100%.

Degraded

Nano Banana 2.1: elevated error rate

– · lasted 1 min · Detected automatically

Failed requests rose noticeably; failed requests are not billed. Success rate during the incident: 78.8%.

Degraded

Grok Imagine Image: elevated error rate

– · lasted 34 min · Detected automatically

Failed requests rose noticeably; failed requests are not billed. Success rate during the incident: 77.1%.

Degraded

Nano Banana 2.1: elevated error rate

– · lasted 48 min · Detected automatically

Failed requests rose noticeably; failed requests are not billed. Success rate during the incident: 82.5%.

Outage

DeepSeek V4.1 Flash: most requests failing

– · lasted 58 min · Detected automatically

Failed requests rose noticeably; failed requests are not billed. Success rate during the incident: 35.9%.

Degraded

GPT Image 2.5 Flare: elevated error rate

– · lasted 1 min · Detected automatically

Failed requests rose noticeably; failed requests are not billed.

Degraded

GPT-6.1 Sol: elevated error rate

– · lasted 53 min · Detected automatically

Failed requests rose noticeably; failed requests are not billed. Success rate during the incident: 88.1%.

How this page is measured

Success rate: each customer request counts once, and a request rescued by a backup line or by another model counts as a success. Not counted: client-side errors (bad parameters, auth, balance and other 4xx), requests refused upstream for their content, requests still in flight, internal tests and banned accounts. A figure appears only once the sample is large enough, and it is truncated, never rounded up.

Degraded: within one hour, at least 3 failed requests and a success rate below 90% (video: 70%); or at least 3 requests with more than half answered by another model; or first-token or generation times well above the model's previous 7 days. Outage: within one hour, at least 3 failed requests and a success rate below 50%, or every line temporarily unavailable. Maintenance: set by us. An incident ends once the model answers on its own again; with no new requests at all, a failure incident ends after 24 hours and an incident of answers from another model after at most 7 days.

Daily bars use UTC days; incident times are shown in your local time zone. The page refreshes every minute.

API endpoint availability 98.26% (last 30 days): the public health-check endpoint is probed every 2 minutes. It shows whether the endpoint is reachable, not whether models answer.

Machine-readable: /status.json · /health